Professional Documents
Culture Documents
12-1995
Michael S. Morgan
Cornell University
Stowe Shoemaker
University of Nevada, Las Vegas
Recommended Citation
Dev, C. S., Morgan, M. S., & Shoemaker, S. (1995). A positioning analysis of hotel brands - Based on travel-manager perceptions.
Cornell Hotel and Restaurant Administration Quarterly, 36(6). 48-55. doi:10.1016/0010-8804(96)81003-2
This Article or Chapter is brought to you for free and open access by the School of Hotel Administration Collection at The Scholarly Commons. It has
been accepted for inclusion in Articles and Chapters by an authorized administrator of The Scholarly Commons. For more information, please contact
hlmdigital@cornell.edu.
A Positioning Analysis of Hotel Brands - Based on Travel-Manager
Perceptions
Abstract
A hotel brands unique selling propositionthe argument it makes to convince travelers to book its hotels
instead of someone elses propertiesis known as its market position. The position comprises the bundle of
attributes that the hotel offers in an effort to meet guests wants and needs.
Keywords
hotel industry, market share, marketing, competitive strategy, branding
Disciplines
Hospitality Administration and Management | Marketing | Tourism and Travel
Comments
Required Publisher Statement
Cornell University. Reprinted with permission. All rights reserved.
Based on Travel-Manager Perceptions position is in the eye of the customer. Identifying that
A Quality of food
B Physical apperanace,
In-room amenities
C Helpful, courteous staff
D Facilities for non-resort meetings
E Overall price value
F Overall average rating,
Facilities for resort meetings
G Ease in arranging individual
travel
H Ease in arranging group travel
Frequent-traveler programs
in the affirmative. The consequence brands in each tier in the data pub cally the changes in market position
of this m ethodology is that the lished in 1990 and then compared over time to determine the extent
sample size from brand to brand in those brands positions in 1991 and to which brands occupy the same
the same years survey shows consid 1992. Each analysis was conducted perceptual space.
erable variation. for 30 hotel brands (10 chains for We applied probabilistic multidi
As a final note on the data, the each of three years in each market mensional scaling (MDS) algorithms
ratings used were means for each segment). We also analyzed the po to derive the coordinates for the
hotel brand. We attempted to obtain sitions of the top two brands in each perceptual map.18 Multidimensional
the entire data set but it was not of the five tiers over the three years scaling is a m ethod of calculating
available. We had no way to deter (another 30-brand analysis). In similarities between objects on a set
mine the level of homogeneity for making the perceptual maps, how of attributes. The calculations result
the sample because variance esti ever, we excluded the luxury tier in coordinates that can be plotted
mates were not given. due to an insufficient num ber of on coordinate axes to form a map.
Analysis. O ur goal was to create respondents. The distances thus calculated give
perceptual maps showing the rela The outcome of our analysis was an indication of the extent to which
tive positions of the various brands perceptual maps that revealed how the respondents view brands as
against each other and to examine each brand was positioned relative similar.
any movement in a brands position to its competitors and how each
during the three years, as viewed by brands position changed over the 18 See: J.O. Ramsay, Some Statistical Approaches
the business travel managers and three years we studied. O ur goal to Multidimensional Scaling,'Journal of the Royal
Statistical Society, Series A,Vol. 145 (1982), pp.
agents. We analyzed the relative was more than simple description, 285-312; and M.L. Davison, Multidimensional
positions of the ten most-used however. We wanted to test statisti Scaling (New York: John W iley & Sons, 1983).
Exhibit 2
Relative market positions (this study):
Upscale tier, 1990
We obtained the MDS algo ties. The coordinates of a brands except if a chain intends to be one
rithms from the com puter program location on X and Y axes reflect place (with one particular com peti
Multiscale, described in the box underlying composites of attitudes tive set) and finds itself at some
on page 55.The algorithms found toward the brands among the survey other place. Perceptual maps can
in Multiscale require that some sort respondents. It is im portant to bear indicate how close one s brand is
of distance m atrix be applied. We in mind that the distances are in to competing brands. Brands that
used a dissimilarities matrix, created psychological space, measured in are positioned relatively far away
using the following Euclidean dis terms of customer perceptions and from each other on the map are
tance metric: preferences rather than on differ interpreted to be less directly com
P ences derived from more objective petitive, while hotel chains that have
<V= ^ <W 2 measures.19
It is also im portant to note that
nearby coordinates are considered to
be strongly competitive with each
k=1
the position maps are essentially other. The map also can identify
for chains i, j, and attributes k=1 p
value-neutral. That is, one spot on open space, which is interpreted to
the map does not inherently have to be an available market niche either
The distances d were then repre be better or worse than another, for repositioning existing operations
sented to the best extent possible on or for a new entrant.
a two-dimensional map. 19 The significance o f psychological positioning A frequent complaint about
Positions of the firms or brands is explored in: George Overstreet, Creating MDS is that the plotted points and
on the perceptual map can assist Value in Oversupplied Markets: The Case of
Charlottesville, Virginia, Hotels, Cornell Hotel
distances are derived purely
managers in identifying potential and Restaurant Administration Quarterly, Vol. 34, algorithmically, w ithout respect to
competitive threats and opportuni No. 5 (October 1993), pp. 84-91. the probability distribution of errors
Exhibit 3
Relative market position (this study):
Upscale tier, 1991 j
Meridien
in the space. We have applied an other. We can also tell whether a to another. Changes in perceptual
algorithm that conducts statistical movement by a brand from year to distances experienced by a brand
tests of significance to ensure the year is significant. The benefit of over time that are statistically signifi
points are, in fact, different. Such a using probabilistic scaling is that it cant we term direction, while
test alleviates the concerns raised by allows one to focus only on position changes that involve m otion that is
the following questions. If the dis differences that are statistically sig not statistically significant we term
tance between two points cannot be nificant. Such an approach clears drift.
perfectly reflected in a reduced two- much of the clutter surrounding
dimensional M DS map, how is the position differences. Cartography
error distributed? Moreover, how The outcome of the calculations Statistical tests allow us to examine
can one tell w hether two points are is a set of points on a map. The lo whether the points on the coordi
really different from one another in cation of each hotel brand is de nate axes shown in Exhibit 2 are at a
a statistically significant way? Even picted according to how customers significant distance from each other.
points that look different may con perceive them on the attribute (Note that Exhibits 2, 3, and 4 can
tain random error that makes them, dimensions in the graph in Exhibit be overlayed on Exhibit 1.) The
in reality, not different. 1. Although the brands are arrayed differences shown in two dimen
Because we have the ability to on coordinate axes, their positions sions may actually be larger in three
test for the significance of differ in relation to the axes (and at dimensions. The circles drawn
ences between points on the per tributes) themselves are less im por around the points on this map de
ceptual map, we can determine tant than their positions in relation pict clusters of points that are not at
which chains have positions that are to one another or the change in a statistically significant distances from
at significant distances from each brands own position from one year each other. So, for example, the
Exhibit 4
Relative market position (this study):
Upscale tier, 1992
Wyndham
Hotels
map of the 1990 data (collected in Changes com ing. The 1991 would never recover from the effects
1989) makes it appear that Stouffers data (collected in 1990) may reflect of war, overbuilding, and recession
position was considerably different the results of the severe recession (Exhibit 4). Ironically, the com peti
from that of M eridien.The statistical that began in 1990. This map (Ex tive clusters found in the 1992 data
tests showed otherwise, as indicated hibit 3) allows two levels of analysis. are similar to those in the baseline
by the circle around those two First, we can once again compare 1990 chart (Exhibit 2). Hilton, Wes
chains. Likewise, there was no sig the chains relative positions against tin, Sheraton, and Hyatt, remain
nificant difference in the perceived each other. We can also compare the fairly consistent. M arriott has moved
positions of W yndham and M eri positions of each chain in 1991 with substantially but is again viewed as
diem However, the distance between its 1990 position. Indeed, there was competitive with Hyatt, albeit not
Wyndham and Stouffer is significant. considerable movement in managers with Westin. Likewise, Radisson is
Viewed another way, the busi- and agents perceptions of the once again viewed as competitive
ness-travel managers and travel chains positions. For instance, M ar with Sheraton and Hilton.
agents in the 1990 survey viewed riott has m oved out of its form er
Hilton, Radisson, and Sheraton as competitive group with Hyatt and Chain Movement
competitive brands whose essential Westin to be more competitive with The perceptual maps give us an idea
positions were not particularly dis Hilton and Sheraton. It is interesting of how individual brands did or did
tinct from one another. Hyatt, M ar that none of those other four chains not attempt to change customers
riott, and Westin form ed another moved substantially from 1990 to perceptions of their positions. The
competitive group, with a position 1991. The movement of M eridien maps capture the movement of indi
significantly different from that of and W yndham was substantial. vidual brands from year to year
Hilton, Radisson, and Sheraton, but 1992 . The data published in 1992 based on customers perceptions of
not from each other. This map pro record the positions hotel brands differences among the chains, which
vides a baseline against which to held in the nightmare year of 1991, the attribute arrows help to illus
compare the other two. when it appeared that the industry trate. The decisions the companies
make to emphasize one or another
discriminating dimension shifts Basic points of probabilistic multidimensional scaling (Multiscale)
their position in the minds of their
customers. (1) Points are located in Euclidean space based on variable ratings (in this case
In the 1990 map, Stouffer is on based on perceptions of product attributes), and fitted to a two-dimensional perceptual
map. Interpoint distances are calculated.
par with Inter-Continental on price (2) The differences between true and fitted interpoint distances (errors) are compared
and product quality, but in 1991, using statistical tests of whether those distances are due to random chance or whether
customers price-value perception of they are statistically significant.
Stouffer moved it away from Inter (3) All distances are considered positive (absolute value). Statistical tests are based
on the assumption that variables have a normal (log-normal) distribution, as depicted in
Continentals competitive set, a the curve below.
movement that reversed in 1992 (4) Any extreme values (outliers) must be transformed. Before applying a Multiscale
following considerable advertising, fitting to these distances, the outliers must be smoothed out so that the final graph